Bridging the Gap: A Machine Learning Approach to Political Knowledge Disparities Between Genders

Stritzel, Laura (2023) Bridging the Gap: A Machine Learning Approach to Political Knowledge Disparities Between Genders. Bachelor thesis, Global Responsibility & Leadership (GRL).

PDF BA_s4553845_LJA_Stritzel.pdf Restricted to Repository staff only Download (2MB)

Abstract

Much research has argued men to be more knowledgeable about politics than women, suggesting the existence of a gender gap in political knowledge. While some scholars have intended to find explanations for this gender gap, examining psychological and social differences between genders, others have framed the issue as an artifact, criticizing measurement properties and survey structures for incorrectly measuring political knowledge. Thus, a common understanding of the phenomenon is still lacking, as it remains unclear to what extent and due to which factors apparent knowledge inequalities exist. Hence, this thesis takes both a quantitative and qualitative approach to reevaluate the persistence of the gender gap in political knowledge from an interdisciplinary perspective. Data from the 2020 American National Election Studies (ANES) is explored, and machine learning algorithms used to test the attempt of correctly classifying an individual's gender, based on the survey responses given. By using this method, the study aims to uncover patterns and draw insights from the survey data which may not be visible to the human eye. The thesis begins by summarizing the current state of knowledge from the existing body of literature. Study methods follow the data science lifecycle in extracting and preparing the data, exploring it by applying chi-squared tests, and performing feature engineering to select the most relevant variables. This statistical analysis demonstrates that significant differences in answer selection are visible between genders. A special focus is put on the exploration of the influence of ‘Don’t know’ answer options on outcomes, as well as answer classifications and techniques of knowledge measurement. A Random Forest model is trained for gender classification, and implications are discussed. Consistent with previous studies, men seem to choose correct answers more frequently than women. However, survey structure and evaluation techniques seem to pose great influence on the outcomes, demonstrating that observations are not necessarily cause of knowledge differences. This thesis contributes to the debate on the gender gap in political knowledge by highlighting that female gender classification on the basis of their survey data is more accurate than male gender classification. This finding shows the practical implications of survey format and measurement biases in influencing both genders differently and therewith especially influencing female response evaluations. These gender inequalities call for a change in survey format and knowledge measurement techniques.

Item Type:	Thesis (Bachelor)
Name supervisor:	Schauble, J.K.
Date Deposited:	12 Sep 2023 09:36
Last Modified:	12 Sep 2023 09:36
URI:	https://campus-fryslan.studenttheses.ub.rug.nl/id/eprint/254

Actions (login required)

View Item